Transcript www.erlang.org

Comparing C++ and Erlang for
Motorola Telecoms Software
Phil Trinder & Henry Nyström
Computer Science Department
Heriot-Watt University, UK
Erlang Training & Consulting
David King
Software & Systems Engineering Research
Motorola Labs, UK
High-Level Techniques for Distributed
Telecoms Software
• EPSRC (UK Govt) Project, Dec 2002 – Feb 2006
• Collaboration between
– Motorola UK Labs
– Heriot-Watt University
High-Level Techniques for Distributed
Telecoms Software
• EPSRC (UK Govt) Project, Dec 2002 – Feb 2006
• Collaboration between
– Motorola UK Labs
– Heriot-Watt University
• Aim: produce scientific evidence that high-level
distributed languages like Erlang or Glasgow distributed
Haskell (GdH) can improve distributed software
robustness and productivity
• Publication: High-Level Distribution for the Rapid
Production of Robust Telecoms Software: comparing C++
and Erlang, Concurrency and Computations: Practice &
Experience (forthcoming).
Erlang Comparisons
A number of sequential comparisons, e.g. Computer
Language Shootout
Very few distributed system comparisons published!
Ulf Wiger [Wiger01] reports
• Erlang systems have between 4 and 10 times less code
than C/C++/Java/PLEX systems
• Similar error rates/line of code
• Similar productivity rates
No direct comparative measurements
Jantsch et al compare 6 languages for hardware description
[JKS+01]
Research Questions: Potential Benefits
RQ1: Can robust, configurable systems be readily developed?
• Resilience to extreme loads
• Availability in the face of hardware & software failures
• Dynamic reconfigurability on available hardware
RQ2: Can productivity and maintainability be improved?
• How do the sizes of the C++ and Erlang components
compare & what language features contribute to size
differential?
Research Questions: Feasibility
High-level distributed languages:
• abrogate control of low-level coordination aspects, so
● RQ3 can the required functionality be specified?
• typically pay space and time penalties for their automatic
coordination management.
● RQ4 can acceptable performance be achieved?
RQ5 What are the costs of interoperating with conventional
technology?
RQ6 Is the technology practical?
Research Strategy
• Reengineer some telecoms application
components in GdH and Erlang
– Dispatch Call Controller [NTK04,NTK05]
– Data Mobility component
• Compare high-level and Java/C++
implementations for
–
–
–
–
Performance
Robustness
Productivity
Impact of programming language constructs
1st Component Engineered:
Data Mobility Component (DM)
•
•
•
•
•
•
Product Component
Communicates with Motorola mobile devices
3000 lines of C++
Uses 18,000 lines of Motorola C library functions
Has a transmitter and a receiver, and 2 message types
Interacts with 5 other components of the system
2nd Component Engineered:
Despatch Call Controller (DCC)
• Handles mobile phone calls
• A process manages each call
• Scalable with multiple servers
Two Erlang DM Implementations
1. Pure Erlang DM
2. Erlang/C DM
reuses some C DM libraries & drivers
Both interoperate with a C test harness
Combine
– Unix Processes
– Erlang processes
– C Threads (Erlang/C DM)
RQ3 Performance 1: Throughput
• Platform: 167MHz, 128Mb Sun Ultra 1, SunOS 5.8
C++ DM
Erlang/C DM
Pure Erlang
DM
480
230
940
Maximum DM Throughput at 100% QoS
RQ3 Performance 1: Throughput
• Platform: 167MHz, 128Mb Sun Ultra 1, SunOS 5.8
C++ DM
Erlang/C DM
Pure Erlang
DM
480
230
940
Maximum DM Throughput at 100% QoS
Pure Erlang DM is twice as fast as C++ DM (better
process management and marshalling)
Erlang/C DM is ½ speed of C++ DM, but still meets
nominal throughput
Performance 2: Round Trip Times
Pure Erlang is approximately 3 times faster
Erlang/C is 26% - 50% slower
C++ DM
Erlang/C DM
Pure Erlang DM
7
ms
6
5
4
3
2
1
0
Query type 1
Query type 2
ms
Broken query
Performance Analysis
Pure Erlang DM is faster due to fast lightweight process
management
Erlang/C is slower due to additional communication to C
components
Performance 3: Memory Residence
Kb
• Erlang DMs use 170%
more memory
• Erlang runtime sys
(ERTS) has a fixed size
• => would be a smaller
% of a larger app.
ERTS
Moto C lib
App C/C++
App Erlang
7000
6000
5000
4000
3000
2000
1000
0
C++ DM
Erlang/C DM
Pure Erlang
DM
RQ1 Robustness 1: Resilience
Erlang A
Throughput
C++ A
Pure Erlang A
1200
(queries/s) 1000
800
600
400
200
Load (queries/s)
24
0
31
0
48
0
94
0
14
00
19
00
47
00
94
00
16
00
0
25
00
0
48
0
31
0
24
0
0
RQ1 Robustness 1: Resilience
Erlang A
Throughput
C++ A
Pure Erlang A
1200
(queries/s) 1000
800
600
400
200
When overloaded:
Load (queries/s)
C++ DM fails catastrophically
Pure Erlang & Erlang/C DMs:
• Throughput degrades
• Never completely fails, handling 48 q/s at peak load
(25000q/s)
- Recovers automatically after load drops
24
0
31
0
48
0
94
0
14
00
19
00
47
00
94
00
16
00
0
25
00
0
48
0
31
0
24
0
0
DCC Resilience
Robustness 2: Availability
Erlang systems
• remain available despite
repeated hardware &
software failures
• performance doesn’t
degrade with repeated
failures
DCC Throughput with Repeated Failures
Robustness 2: Availability
Erlang Systems resists the
simultaneous failure of
multiple components
When more components fail
throughput drops lower &
recovery takes longer
5-processor DCC with Multiple Failures
Robustness 3: Dynamic Configurability
Erlang Systems dynamically
adapt to the available
hardware resources.
5 processor system:
– remove a processor 4 times
– add a processor 4 times
DCC Throughput
with Varying Numbers of Processors
RQ2: Productivity & Maintainability
Shorter programs are
• Faster to develop
• Contain fewer errors [Wiger01]
• Easier to maintain
The metric we use is Source Lines Of Code (SLOC)
Productivity: Source Code Sizes
Lang.
C/C++
C++
3101
Erl./C
247
Erlang
Erlang Total
Lang. C++
3101
C++
616
863
Erl.
398
398
IDL
Erl.
14.8K 83
14.9K
4882
DCC Implementations
DM Implementations
Erlang DCC and DM are less than 1/3rd of size of C++ impl.
Consistent with Wiger & folklore
Total
4882
Productivity: DM Source Code Sizes
C++/C
Erlang
3500
3000
2500
2000
1500
1000
500
0
C++ DM
Erlang/C DM
Pure Erlang DM
• Erlang/C DM is 1/3rd of the size of the C++ DM
• Pure Erlang DM is 1/7th of the size of the C++ DM
• Erlang/C DM is 1/18th of the size of the C++ DM +
libraries
Reasons for difference in Code Size
• Erlang programmers can
– rely on fault tolerance and code for the successful
case (27% of C++ DM code is defensive)
– and have
• automatic memory management (11% of C++ DM code)
• high-level communication (23% of C++ DM code)
• Telecom design pattern libraries
DM Code Breakdown
100%
90%
80%
Defensive
Defines
Includes
Type Delcarations
Communication
Memory Management
Process Management
App
70%
60%
50%
40%
30%
20%
10%
0%
C++ A
Moto Clib Erlang/C
Erlang
Code Difference Example
C++
void DataMobilityRxProcessor::processUnsupVer(void)
{
MSG_PTR
msg_buf_ptr;
MM_DEVICE_INFO_MSG *msg_ptr;
RETURN_STATUS
UINT16
ret_status;
msg_size;
// Determine size of ici message
msg_size = sizeof( MM_DEVICE_INFO_MSG);
// Create ICI message object to send to DMTX so it sends a Device Info
// message to VLR and HLR clients
IciMsg ici_msg_object( MM_DEVICE_INFO_OPC, ICI_DMTX_TASK_ID, msg_size);
// Retrieve ICI message buffer pointer
msg_buf_ptr = ici_msg_object.getIciMsgBufPtr();
// Typecast pointer from (void *) to (MM_DEVICE_INFO_MSG *)
msg_ptr = (MM_DEVICE_INFO_MSG *)msg_buf_ptr;
// Populate message buffer
SET_MM_DEVICE_INFO_DEVICE_TYPE( msg_ptr, SERVER);
SET_MM_DEVICE_INFO_NUM_VER_SUPPORTED( msg_ptr, NUM_VER_SUPPORTED);
SET_MM_DEVICE_INFO_FIRST_SUP_PROTO_VERS( msg_ptr, PROTO_VERSION_ONE);
// Send message to the DMTX task
ret_status = m_ici_io_ptr->send(&ici_msg_object);
// Check that message was sent successfully
if (ret_status != SUCCESS)
{
// Report problem when sending ICI message
sz_err_msg( MAJOR, SZ_ERR_MSG_ERR_OPCODE, __FILE__, __LINE__,
"DataMobilityRxProcessor processUnsupVer: failure sending "
" device info message to DMTX");
}
}
Erlang
sz_dme_dmtx:cast(device_info)
Erlang DCC Reusability
Part
SLOC
No. Modules
Percentage
Reusable
Platform
2994
26
61%
Specific
Services
147
1
3%
Testing/Stat.s
1741
11
36%
Considerable potential for reuse
Summary
• Investigated high-level distributed language
technology for telecoms software
• Reengineered two telecoms components in Erlang
• Measured & compared the Erlang & C++ components
RQ1: Robust & Configurable Systems
• Improved resilience:
– Erlang DM and DCC sustain throughput at extreme loads
– Automatically recover when load drops
– C++ DM fails catastrophically (predict C++/CORBA DCC would)
• Improved availability:
– Erlang DCC recovers from repeated & multiple failures
– Predict C++/CORBA DCC would fail catastrophically
• Dynamic Reconfiguration
– Erlang DCC can be dynamically reconfigured to available
hardware
– C++/CORBA DCC can also be dynamically reconfigured using
CORBA
• Potential for hot-code loading (not demonstrated)
RQ2: Productivity & Maintainability
• Erlang DM and DCC:
– Less than 1/3rd size of C++ implementation
– Erlang DM 1/18th size of C++ DM with libraries
– Good reusability
• Reasons:
– Code for successful case – saves 27%
– Automatic memory management – saves 11%
– High-level communications – saves 23%
– Telecom design pattern libraries
RQ3: Distributed Functionality
• Even though Erlang abstracts over low-level
coordination, the required DM and DCC functionality
is readily specified.
RQ4: Performance
• Time:
– Max. throughput at 100% QoS:
• Pure Erlang DM is twice as fast as C++ DM
• Erlang/C is ½ as fast as C++ DM , but still exceeds
throughput requirements
– Roundtrip times
• Pure Erlang DM is three times as fast as C++ DM
• Erlang/C is between 26% and 50% slower as C++ DM
• Space:
– Pure Erlang and Erlang/C both have 170% greater
memory residency due to (fixed size) 5Mb runtime
system
RQ5: Interoperation Costs
• Erlang DMs interoperate with a C test harness, and
Erlang/C DM incorporates C drivers & library
functions.
• Costs
– Low space cost: an additional 15% residency
– High time cost:
• Erlang/C roundtrip times up to 6 times pure Erlang
• Erlang/C max. throughput ¼ of pure Erlang
• Potential for incremental re-engineering of large
systems
RQ6: Pragmatics
• Erlang is available on several HW/OS platforms,
including the Sun/Solaris DM product platform
• Well supported with training, consultancy, libraries
etc.
Conclusions
• Erlang offers robustness & productivity benefits for
distributed telecoms software (RQs 1 & 2)
• High-level distributed languages like Erlang can
deliver the required telecoms functionality and
performance (RQs 3 & 4)
• Erlang can interoperate with existing technologies
and meets pragmatic requirements (RQs 5 and 6)
Further Information
Web sites/Seminars
Erlang Site: www.erlang.org/
Project Site: www.macs.hw.ac.uk/~dsg/telecoms/
References
[JKS+01] Jantsch A. Kumar S. Sander I. Svantesson B. Oberg J. Hemanic A. Ellervee P. O’Nils
M. Comparison of Six Languages for System Level Descriptions of Telecoms Systems,
Electronic Chips and System Design, pp 181-192, Kluwer, 2001.
[NTK04] Nystrom J.H. Trinder P.W. King D.J. Evaluating Erlang for Robust
Telecoms Software S3S’04, Chicago, July 2004.
[NTK05] Nystrom J.H. Trinder P.W. King D.J. Are High-Level languages suitable for
Robust Telecoms Software? SafeComp’05, Fredrikstad, Norway, Sept. 2005.
[Wiger01] Ulf Wiger, Workshop on Formal Design of Safety Critical Embedded
Systems, Munich, March 2001 http://www.erlang.se/publications/Ulf_Wiger.pdf